The invention pertains to network data transmission controls. More particularly, the invention relates to systems for minimizing the effects of packet flooding attacks directed against computers or routers connected to a network.
Various types of systems have been developed for handling unwanted network data transmission incorporating a number of different technologies. U.S. Pat. No. 5,581,559 issued to Crayford et al. discloses a method that verifies the integrity of data transmitted over a network by comparing the destination address for a data packet with end station addresses stored on network repeaters. Where the destination address fails to match the stored end station addresses, the data packet will be disrupted.
U.S. Pat. No. 6,044,402 issued to Jacobson et al., describes a system in which the only data packets that are transmitted between source and destination network addresses are those that satisfy the blocking policies stored by the blocking data structure. Thus only, xe2x80x9cpre-approvedxe2x80x9d data can flow through such a control mechanism. U.S. Pat. No. 5,455,865, issued to Perlman discloses a system that relies upon a stored list of acceptable packet identifiers at each node in the network. U.S. Pat. No. 5,353,353 issued to Vijeh et al. describes a system that determines the acceptability of data packets based upon a destination address/source address match and will disrupt any packet not satisfying these criteria. U.S. Pat. No. 5,850,515 issued to Lo et al. discloses a system that uses source and destination address matching to determine if packets should be transmitted to an end station or the end station disabled from participating in the network. It also employs a system where an end station can be disabled by a program that determines that a certain number of unauthorized packets have been detected. While other variations exist, the above-described designs for handling unwanted network data transmissions are typical of those encountered in the prior art.
The primary objective of the present invention is to defend against xe2x80x9cpacket flooding attacksxe2x80x9d in which an attacker tries to use up all the bandwidth to the victim by sending data of little or no value (at least to the victim), thereby making more valuable communication with the victim slow or unreliable. A secondary objective is to defend against a related class of attacks in which the attacker tries to use up some other resource by sending more requests of some particular type to the victim than the victim can handle.
One way to view all these attacks is that a resource is being allocated in an unfair way. Well-behaved users request reasonable amounts, while attackers request unreasonable amounts. The most straight-forward allocation mechanism, which might be called xe2x80x9cfirst come first servedxe2x80x9d, ends up allocating almost all of the resource to the attackers. A more xe2x80x9cfairxe2x80x9d allocation would reduce the impact of an attacker to that of a normal user.
There are two obvious impediments to the xe2x80x9cfair servicexe2x80x9d goal above. One is lack of a reliable way to associate incoming packets with those users among whom bandwidth should be fairly allocated. The other is lack of control over what packets arrive. The solution described here to both of these problems requires help from the routers that forward packets to the victim.
The defense is distributed among cooperating sites and routers. A set of transitively connected cooperating machines is called a xe2x80x9ccooperating neighborhoodxe2x80x9d. The quality of the defense is related to the size of the cooperating neighborhood, a larger neighborhood providing better defense. Within the neighborhood it is possible to trace the forwarding path of packets. The association of packets with the xe2x80x9cusersxe2x80x9d is approximated by associating packets with xe2x80x9cplacesxe2x80x9d in the cooperating neighborhood from which those packets are forwarded. That is, service will be allocated in a fair (or otherwise reasonable) manner among these places. A xe2x80x9cplacexe2x80x9d in this sense is typically a particular interface from which a packet arrived at a cooperating router.
One such place is likely to be shared by many actual users. An attack will deny service to those users sharing the same place. The advantage of a large number of such places is that each place is shared by fewer users, so an attack will deny service to fewer users. It is advantageous to a user who wants to communicate with a particular machine, to be in the cooperating neighborhood of that machine, since no attacker from another machine can deny him service. Conversely, an attacker wishing to deny service to as many users as possible prefers to share an entry point into the cooperating neighborhood with as many users as possible.
Routers will supply data about the forwarding path of the packets that arrive at a site. The site can use this data to allocate service as described above among the packets that arrive. This would solve the problem of unfair service if the packets that arrived were a fair sample of those that were sent to the site. This may not be the case, however, if routers are unable to forward all the packets they receive. To some extent fair service is limited by network topology, i.e., too many legitimate users trying to share parts of the same path will inevitably suffer relative to users of uncrowded paths. However another potential cause for this problem is a flooding attack against a router. That problem is solved by letting routers allocate their services in a similar way to that described above for sites. That is, they allocate the limited resource of forwarding bandwidth along any given output in a reasonable way among some set of places in the cooperating neighborhood.
The final step in the defense is that cooperating routers will limit the rate at which they forward packets to places that so request. This may not be essential in the allocation of service, but it is useful for limiting the bandwidth used by xe2x80x9cunwantedxe2x80x9d packets. The rate-limiting request is to be made when a site detects a high rate of unwanted packets coming from one place. This helps the site because it no longer has to process as many unwanted packets. It helps the network by freeing some of the bandwidth for other use.
Even if the traffic is not reduced, the distinction between xe2x80x9cwantedxe2x80x9d and xe2x80x9cunwantedxe2x80x9d packets plays an important role in xe2x80x9creasonablexe2x80x9d allocation. For a site there are normally some packets (in fact, the great majority) that are expected in a very strong sense. It is reasonable to process these at the highest possible rate. All other packets are not exactly unwanted, but the site is willing to process them at only a limited rate. A reasonable approach is to schedule these as described above (using the places from which they were forwarded) at a limited rate, and regard as xe2x80x9cunwantedxe2x80x9d those that end up being significantly delayed (or discarded).
The present invention addresses many of the deficiencies of prior network defense systems and satisfies all of the objectives described above.
A packet flooding defense system for a network providing the desired features may be constructed from the following components. The network includes a plurality of host computers, routers, communication lines and transmitted data packets. Means are provided for classifying data packets received at a host computer as are means for associating a maximum acceptable processing rate with each class of data packet received at the computer. Means are also provided for the computer to find information for packets it receives regarding the path by which the packets came to the computer. Thus, the computer can use the information to allocate the processing rate available for packets of each class in a desired way.
In another variant, a packet flooding defense system for a network including a plurality of host computers, routers, communication lines and transmitted data packets includes means for classifying data packets received at a host computer and means for associating a maximum acceptable processing rate with each class of data packet received at the computer. Means are provided for the computer to determine the rate at which data packets of each class are transmitted from a router to the computer as are means for the router to receive information regarding maximum acceptable transmission rate for data packets being transmitted to the computer. Means are provided for the router to control the rate of transmission of data packets from the router to the computer. Thus, the rate of data packet transmissions received at the computer is kept below the maximum acceptable processing rate for each data packet class by the control of the rate of transmission of data packets from the router, thereby freeing a portion of the network providing data packet transmission to the computer.
In yet another variant, the router is capable of receiving information regarding maximum acceptable transmission rate for each class of data packet being transmitted to the computer and the router is capable of controlling the rate of transmission of each class of data packets to the computer.
In still another variant, a packet flooding defense system for a network including a plurality of host computers, routers, communication lines and transmitted data packets includes means for classifying data packets received at a router and means for associating a maximum acceptable transmission rate with each class of data packet received at the router. Means are provided for the router to find information for packets it receives regarding the path by which the packets came to the router. Thus, the router can use the information to allocate the transmission rate for each class in a desired way.
In a further variant of the invention, a packet flooding defense system for a network including a plurality of host computers, routers, communication lines and transmitted data packets includes means for classifying data packets received at a first router and means for associating a maximum acceptable transmission rate with each class of data packet received at the first router. Means are provided for the first router to determine the rate at which data packets of each class are transmitted from a second router to the first router as are means for the second router to receive information regarding maximum acceptable transmission rate for data packets being transmitted to the first router. Means are provided for the second router to control the rate of transmission of data packets from the second router to the first router. Thus, the rate of data packet transmissions received at the first router is kept below the maximum acceptable transmission rate for each data packet class by the control of the rate of transmission of data packets from the second router, thereby freeing a portion of the network providing data packet transmission to the first router.
In yet a further variant, the second router is capable of receiving information regarding maximum acceptable transmission rate for each class of data packet being transmitted to the first router and the second router is capable of controlling the rate of transmission of each class of data packets to the first router.
In another variant, a packet flooding defense system for a network including a plurality of host computers, routers, communication lines and transmitted data packets includes at least one firewall. The firewall includes hardware and software serving to control packet transmission between the network and a host computer connected to an internal network. Means are provided for classifying data packets received at the firewall as are means for associating a maximum acceptable transmission rate with each class of data packet received at the firewall. Means are provided for the firewall to find information for packets it receives regarding the path by which the packets came to the firewall. Thus, the firewall can use the information to allocate the transmission rate for each class in a desired way.
In still another variant of the invention, a packet flooding defense system for a network including a plurality of host computers, routers, communication lines and transmitted data packets includes at least one firewall. The firewall includes hardware and software serving to control packet transmission between the network and a host computer connected to an internal network and means for classifying data packets received at the firewall. Means are provided for associating a maximum acceptable transmission rate with each class of data packet received at the firewall as are means for the firewall to determine the rate at which data packets of each class are transmitted from a router to the firewall. Means are provided for the router to receive information regarding maximum acceptable transmission rate for data packets being transmitted to the firewall as are means for the router to control the rate of transmission of data packets from the router to the firewall. Thus, the rate of data packet transmissions received at the firewall is kept below the maximum acceptable transmission rate for each data packet class by the control of the rate of transmission of data packets from the router, thereby freeing a portion of the network providing data packet transmission to the firewall.
In a final variant of the invention, the router is capable of receiving information regarding maximum acceptable transmission rate for each class of data packet being transmitted to the firewall and the router is capable of controlling the rate of transmission of each class of data packets to the firewall.
An appreciation of the other aims and objectives of the present invention and an understanding of it may be achieved by referring to the accompanying drawings and the detailed description of a preferred embodiment.